Diagnostic BMI and Induction Therapy Weight Change Associated With Hepatotoxicity in Pediatric Patients With Acute Lymphoblastic Leukemia Authors: Emily J. Mason, John P. Woodhouse, Jeremy M. Schraw, Joanna S. Yi, Karen R. Rabin, M. Monica Gramatges, Mark C. Zobeck, Baojiang Chen, Melissa B. Harrell, Olga A. Taylor, Michael E. Scheurer, Philip J. Lupo, Van Huynh, Steven D. Mittelman, Etan Orgel, Austin Brown Published: 2026 ABSTRACT ABSTRACTBackgroundHepatotoxicity is a common treatment‐related complication in pediatric acute lymphoblastic leukemia (ALL). Excess adiposity at diagnosis has been identified as a risk factor, but the impact of weight change during therapy has not been examined.MethodsWe included patients aged 2–20 years newly diagnosed with ALL between 2005 and 2021. Hepatotoxicity was defined as: (1) transaminitis (ALT/AST > 10× the upper limit of normal), or (2) conjugated hyperbilirubinemia (cbili; > 3.0 mg/dL during induction/> 2.0 mg/dL post‐induction). Youden's J was used to identify optimal BMI z‐score thresholds. Cox proportional hazards regression was used to assess associations between hepatotoxicity during treatment and diagnostic BMI and post‐induction hepatotoxicity and weight change during induction. Associations were adjusted for age, race, ethnicity, and treatment risk stratification. Patients were censored at qualifying event or 1200 days post‐diagnosis.ResultsAmong 1070 patients (median age 6 years), 45.8% developed transaminitis and 7.6% cbili. Optimal BMI z‐score cut points were −1.5 and 1.5 for transaminitis and 1.5 for cbili. A BMI z‐score ≤ −1.5 was associated with reduced transaminitis (HR = 0.58, 95% CI: 0.39–0.86; p = 0.006), while a z‐score ≥ 1.5 was associated with increased transaminitis (HR = 1.32, 95% CI: 1.06–1.65; p = 0.01). A BMI z‐score ≥ 1.5 was also associated with increased cbili (HR = 1.97, 95% CI: 1.23–3.18; p = 0.01). Weight change was not associated with transaminitis, but weight loss during induction was associated with an increase in hazard of post‐induction cbili (HR = 1.08, 95% CI: 1.04–1.11; p < 0.0001).ConclusionDiagnostic BMI z‐score and weight loss during induction were associated with increased hepatotoxicity risk, highlighting potential targets for improved risk stratification and hepatoprotective interventions. Introduction Acute lymphoblastic leukemia (ALL) is the most common pediatric cancer in the United States, with approximately 3100 cases diagnosed each year [1, 2]. While current treatment regimens have resulted in long‐term survival rates approaching 90% [3, 4], approximately 30%–50% of ALL patients experience at least one serious therapy‐related adverse event during treatment [1]. In particular, treatment‐associated hepatotoxicity (TAH) is one of the most frequent non‐hematologic toxicities observed in pediatric patients with ALL due to the use of hepatotoxic agents such as asparaginase and methotrexate. Importantly, severe TAH can require treatment modifications, potentially compromising the efficacy of treatment and resulting in poorer overall and event‐free survival [5]. Despite being a significant cause of morbidity in children with ALL, risk factors for TAH are poorly understood. Elevated body mass index (BMI) at ALL diagnosis has been associated with an increased risk of TAH in children with ALL [6, 7, 8, 9, 10, 11, 12]. Prior studies have generally examined BMI as a categorical variable (e.g., underweight, healthy weight, overweight, obese) based on general population thresholds [6, 9, 10, 11, 13, 14]. While valuable for population‐level analyses, these categories may not reflect risk thresholds for adverse events in pediatric ALL. Additionally, ALL chemotherapy is associated with significant weight gain, particularly during the initial induction phase [15, 16]. Thus, while BMI at diagnosis has been linked to toxicity risk, a BMI measurement at baseline may not accurately reflect risk throughout the course of therapy. Therefore, this study aimed to evaluate associations of diagnostic BMI and post‐diagnosis weight change with TAH in a cohort of children with ALL. Methods This retrospective chart review utilized data from patients newly diagnosed with ALL at Texas Children's Hospital in Houston, TX between 2006 and 2021. Eligible patients were aged 2–20 years at diagnosis and treated for B‐ALL or T‐ALL. Patients with down syndrome or other conditions that limit the use of BMI growth curves were excluded. Clinical and demographic data including race, ethnicity, age at diagnosis, ALL immunophenotype, and treatment risk group were abstracted from electronic health records of eligible participants. All patients received a 4‐drug induction except NCI standard risk (SR) B‐ALL patients without steroid pretreatment or extramedullary involvement and post‐induction risk stratification was determined by NCI risk stratification, leukemia cytogenetics and disease response. Some patients with BMI > 95th percentile were dose capped for asparaginase at 3750 units/dose; this practice was implemented in approximately 2020/2021 at Texas Children's Hospital, so only a small subset of more recently diagnosed patients would have been subject to dose capping. Height (cm) and weight (kg) were manually abstracted from medical records of directly observed weight and height at diagnosis and the start of each phase of therapy. BMI was calculated from the ratio of weight (kg) to the square of height (m). BMI z‐scores were calculated using Centers for Disease Control and Prevention (CDC) recommended age‐ and sex‐specific growth curves [17, 18, 19]. Weight change was expressed in kg and then converted into percent change. Percentage of diagnostic body weight change during induction was calculated by the following equation: [consolidation day 1 weight (kg)−dx weight (kg)]/[dx weight (kg)] × 100. Outcomes Patients were followed post‐diagnosis for the development of TAH events. Patients were followed for 1200 days, or 3.25 years, post‐diagnosis and censored at end of treatment, transfer, time of relapse, time of stem cell transplant, death, or loss to follow‐up. Liver labs, typically collected at the beginning of each phase of therapy and each cycle of maintenance with additional labs collected at the discretion of the treating physician, were evaluated to define TAH as: (1) conjugated hyperbilirubinemia if conjugated bilirubin > 3 mg/dL during induction or > 2 mg/dL post‐induction, and (2) transaminitis defined as an alanine transaminase (ALT) or aspartate aminotransferase (AST) > 10× the upper limit of normal (ULN). These cutoffs are in line with prior publications and defined by clinically relevant thresholds during leukemia treatment that prompt dose modification in most contemporary ALL treatment protocols [6, 13, 20]. Statistical Analysis Statistical analysis was performed in R version 4.3.1 and the survival package (v3.8–3) [21] with statistical significance defined as p < 0.05. Descriptive statistics, including counts and percentage of the total for categorical variables and median and range for continuous variables, were calculated for the entire cohort. Transaminitis and conjugated hyperbilirubinemia outcomes were analyzed separately. Baseline BMI Models First, we evaluated the relationship between BMI at diagnosis and the incidence of TAH during any phase of treatment from induction day two through the end of treatment or time of censoring. The optimal BMI z‐score cut points were defined for each outcome using Youden's J to determine inflection points. Youden's J indicates the point(s) of maximum difference(s) between sensitivity and 1‐specificity and is often used to determine ideal cut points [22]. Cox proportional hazards regression was employed to evaluate the association between time to TAH and baseline BMI z‐score categories determined using thresholds described above. Resulting hazard ratios (HRs) and 95% confidence intervals (CIs) were adjusted for age at diagnosis (continuous), race (White/Black/Asian/Native American/other or unknown), ethnicity (Hispanic/non‐Hispanic), and treatment risk‐stratification (B‐ALL SR‐low or average risk/B‐ALL SR‐high/B‐ALL high risk (HR)/T‐ALL). These were selected a priori due to their reported association with TAH in the existing literature [6, 10, 11, 12, 13, 23, 24, 25, 26]. As age is a well‐known prognostic factor associated with both weight loss during induction and hepatotoxicity [11, 27, 28], age stratified models were examined using age groups of 2–9 and 10+ years. Proportional hazards assumptions were assessed using Schoenfeld residual tests and visual inspection of complementary log–log survival plots stratified by weight‐change quartiles. Weight Change Analysis Second, weight change trajectories throughout treatment were plotted for the entire sample and stratified on diagnostic BMI z‐score category. The percentage change from diagnosis was calculated and the mean percent change in each BMI group was used to plot weight change during each phase of therapy. Cox proportional hazards regression was used to examine the association between weight change from diagnosis to consolidation day 1 and post‐induction TAH, controlling for age, diagnostic BMI z‐score category, race, ethnicity, and treatment protocol. Age‐stratified models were examined to evaluate differences between age groups. Weight change was also evaluated as a categorical variable: ≥ 5% weight loss, 0%–5% weight loss, 0%–5% weight gain, and ≥ 5% weight gain in primary post‐induction models. Stratified cumulative incidence curves over the primary follow up time of consolidation day 1 to 1200 days post diagnosis were plotted by binary weight change category (gain/loss) if weight change was significant in the model. Significant differences were assessed using Gray's test. Sensitivity Analysis This study includes patients diagnosed over 15 years (2006–2021). During this time period, there have been changes in treatment protocols, asparaginase formulations, supportive care practices, and hepatotoxicity monitoring. In order to account for some of these potential changes, incidence of hepatotoxicity by diagnostic era was evaluated and cox regression models were stratified based on the median year of diagnosis and used to assess whether associations were consistent across the study period. Results A total of 1070 eligible patients (age 2–20 years old) were diagnosed with B‐ or T‐ALL at a median age at diagnosis of 6 years. The cohort was 56.2% male, 54.4% Hispanic, 74.0% White, and 86.9% B‐ALL. Around 30% of the cohort was treated on the AALL1131 protocol, followed by AALL0932 (23.4%) and AALL0331 (15.3%) (Table S1). Most of the cohort met CDC definitions for healthy weight (−1.65 < BMI z score < 1.65) at diagnosis (56.2%), while 16.9% were overweight (1.0 ≤ BMI z‐score < 1.65), 18.3% were obese (BMI z‐score ≥ 1.65), and 8.6% were underweight (BMI z‐score ≤ −1.65). The median BMI z‐score was 0.5 with a range between −6.7 and 6.0. Overall, 45.7% of patients (n = 489) experienced transaminitis during follow‐up while 7.6% (n = 81) experienced conjugated hyperbilirubinemia (Table 1). TABLE 1Demographic characteristics of children with ALL.Demographics n = 1070Age, years, median (range)6 (2–20)SexMale601 (56.2%)Female469 (43.8%)BMI categoryObese188 (18.3%)Overweight174 (16.9%)Normal weight579 (56.2%)Underweight89 (8.6%)RaceWhite789 (74.0%)Black85 (8.0%)Asian76 (7.1%)Native American69 (6.5%)Other/unknown47 (4.4%)EthnicityHispanic582 (54.4%)Non‐Hispanic487 (45.6%)Treatment protocolB‐ALL SR‐low/avg432 (40.4%)B‐ALL SR‐high108 (10.1%)B‐ALL HR390 (36.4%)T‐ALL140 (13.1%)Transaminitis a Yes489 (45.7%)No581 (54.3%)Conjugated hyperbilirubinemia b Yes81 (7.6%)No989 (92.4%)Abbreviations: ALT, alanine transaminase; AST, aspartate aminotransferase; BMI, Body mass index; c.bili, conjugated hyperbilirubinemia; HR, NCI high risk; SR, NCI standard risk. a Transaminitis = ALT or AST > 10 × upper limit of normal. b Conjugated hyperbilirubinemia = conjugated bilirubin > 3.0 mg/dL during induction or > 2.0 mg/dL post‐induction. Demographic characteristics of children with ALL. Abbreviations: ALT, alanine transaminase; AST, aspartate aminotransferase; BMI, Body mass index; c.bili, conjugated hyperbilirubinemia; HR, NCI high risk; SR, NCI standard risk. Transaminitis = ALT or AST > 10 × upper limit of normal. Conjugated hyperbilirubinemia = conjugated bilirubin > 3.0 mg/dL during induction or > 2.0 mg/dL post‐induction. Diagnostic BMI Analysis BMI Cut Point Identification A total of 1003 children had height and weight available at diagnosis and were included in the baseline BMI models. Diagnostic BMI z‐score inflection points for transaminitis were identified at z‐scores of approximately −1.5 and 1.5 (Figure S1A). Conversely, a cut point analysis examining conjugated hyperbilirubinemia indicated only one diagnostic BMI z‐score inflection point at z≈1.5 (Figure S1B). Diagnostic BMI Models In the diagnostic BMI z‐score transaminitis multivariable Cox regression model using a categorical BMI variable, diagnostic BMI z‐score of ≥ 1.5 was associated with an increased hazard of transaminitis (HR: 1.32, 95% CI: 1.06–1.65; p = 0.01), while diagnostic BMI z‐score of ≤ −1.5 was associated with a decreased hazard (HR: 0.58, 95% CI: 0.39–0.86; p = 0.006) compared to normal weight children (−1.5 < BMI z‐score < 1.5) (Table 2). Older age at diagnosis and Hispanic ethnicity were also independently associated with an increased hazard of transaminitis. Age stratified models indicated that diagnostic BMI z‐score ≥ 1.5 was significantly associated with an increased hazard of transaminitis for children aged 2–9 at diagnosis, while diagnostic BMI z‐score ≤ −1.5 was associated with a decreased risk of transaminitis for children aged 10 or older at diagnosis (Table S2). For conjugated hyperbilirubinemia, BMI z‐score ≥ 1.5 increased the hazard of conjugated hyperbilirubinemia (HR = 1.97, 95% CI: 1.23–3.18; p = 0.02), as did older age at diagnosis (Table 3). This effect was consistent for children over 10 years old at diagnosis, but diagnostic BMI was not significantly associated with conjugated hyperbilirubinemia for children aged 2–9 years (Table S3). TABLE 2Cox regression model of diagnostic BMI and transaminitis throughout treatment.HR a (95% CI) p BMI category−1.5 < BMI z < 1.5REFBMI z ≥ 1.51.32 (1.06–1.65)0.01* BMI z ≤ −1.50.58 (0.39–0.86)0.006** Age1.03 (1.01–1.06)0.02* Hispanic ethnicity1.46 (1.17–1.81)< 0.001*** a HR additionally adjusted for race and treatment risk stratification.*** p < 0.001.** p < 0.01.* p < 0.05. Cox regression model of diagnostic BMI and transaminitis throughout treatment. HR additionally adjusted for race and treatment risk stratification. p < 0.001. p < 0.01. p < 0.05. TABLE 3Cox regression model of diagnostic BMI and conjugated hyperbilirubinemia throughout treatment.HR a (95% CI) p Hispanic ethnicity1.64 (0.92–2.94)0.09Age at diagnosis1.15 (1.08–1.23)< 0.001*** BMI ≥ 1.51.97 (1.23–3.18)0.015* a HR additionally adjusted for race and treatment risk stratification.*** p < 0.001.* p < 0.05. Cox regression model of diagnostic BMI and conjugated hyperbilirubinemia throughout treatment. HR additionally adjusted for race and treatment risk stratification. p < 0.001. p < 0.05. Weight Change Analysis Weight Trajectories During induction, 66% of children gained weight, with 48% gaining ≥ 5% of diagnostic weight. Overall, 33% of children lost weight during induction, with 19% of the overall sample losing ≥ 5% of diagnostic weight and 14% losing < 5% of their diagnostic weight. On average, children with ALL gained 5% of their diagnostic weight between diagnosis and the start of consolidation and remained 6% above their diagnostic weights at the start of maintenance therapy (Figure 1A). However, weight change trajectories differed by diagnostic BMI z‐score category. For example, children with a BMI z‐score < 1.5 gained an average of 7% of their diagnostic weight during induction therapy, but children with a BMI z score ≥ 1.5 at diagnosis lost an average of 2.25% of their diagnostic weight during induction therapy before returning to baseline at the start of maintenance (Figure 1B). Pre‐maintenance weight change trajectories for the full sample (A) and stratified by diagnostic BMI (B). DX = diagnosis, IM1 = interim maintenance 1, DI = delayed intensification, IM2 = interim maintenance 2. Weight Change Models For the primary endpoint of post‐induction TAH, there were 90 children with missing baseline BMI, consolidation weight, or post‐induction lab values, resulting in an analyzable sample size of 980. Among these patients, 384 patients experienced post‐induction transaminitis, and 53 experienced post‐induction conjugated hyperbilirubinemia. After adding a percent of weight change from diagnosis to consolidation day 1 variable (continuous) to the transaminitis multivariable Cox regression model, weight change during induction was not statistically significant in the overall model (Table S4), nor in age stratified models (Table S5). Secondary multivariable Cox models employing a categorical weight change variable did not indicate any significant differences between weight change categories, and p for trend was not significant (Table S6). Weight loss during induction therapy was associated with an increased hazard of post‐induction conjugated hyperbilirubinemia (HR = 1.08, 95% CI: 1.04–1.11; p < 0.0001) in a multivariable model. Once weight change was added to the model, BMI z‐score at diagnosis was no longer a significant predictor of conjugated hyperbilirubinemia (HR = 1.29, 95% CI: 0.71–2.35; p = 0.40) (Table 4). There appears to be an increase in the number of post‐induction conjugated hyperbilirubinemia events among those who lost any weight during induction compared to patients who maintained or gained weight (p < 0.0001; Figure 2). Notably, weight loss during induction was associated with the hazard of post‐induction conjugated hyperbilirubinemia across age groups (Table S7). Models employing a categorical weight loss variable indicated significantly increased hazard for both 0%–5% and ≥ 5% weight loss compared to the 48% of children who experienced ≥ 5% weight gain. Children who lost 0%–5% and ≥ 5% of their diagnostic weight had 6.05 (95% CI: 2.05–17.86; p = 0.001) and 6.49 (95% CI: 2.28–18.47; p < 0.001) times the hazard of post‐induction conjugated hyperbilirubinemia, respectively. P for trend was significant (p < 0.001) (Table S8). TABLE 4Cox regression model of induction weight change (%) and post‐induction conjugated hyperbilirubinemia.HR a (95% CI) p Induction weight loss (%)1.08 (1.04–1.11)< 0.001*** Hispanic ethnicity1.26 (0.66–2.41)0.48Age at diagnosis1.14 (1.05–1.23)0.001** BMI ≥ 1.51.29 (0.71–2.37)0.40 a HR additionally adjusted for race and treatment risk stratification.*** p < 0.001.** p < 0.01. Cox regression model of induction weight change (%) and post‐induction conjugated hyperbilirubinemia. HR additionally adjusted for race and treatment risk stratification. p < 0.001. p < 0.01. Cumulative incidence of post‐induction conjugated hyperbilirubinemia by induction weight change group. C.bili = conjugated bilirubin. Sensitivity Analysis The median year of diagnosis in the sample was 2014, creating two diagnosis eras (2006‐2013/2014‐2021). The incidence of transaminitis did not significantly change between these time periods, though it was nominally reduced in children diagnosed between 2014 and 2021 (48.6% vs. 43.5%; p = 0.11). The incidence of conjugated hyperbilirubinemia increased over time (5.6% vs. 9.0%), and this effect was borderline statistically significant (p = 0.05). Stratified models adjusted for race, ethnicity, risk group, and age demonstrated that BMI and transaminitis were significantly associated in 2014–2021, but this effect was attenuated in earlier years (Table S9). As in primary and secondary analyses, there was no association between induction weight change and post induction transaminitis (Table S10). For conjugated hyperbilirubinemia, the magnitude and direction of the association between diagnostic BMI and hepatotoxicity were consistent across time periods, though only 2014–2021 reached statistical significance (Table S11). Induction weight change was significantly associated with conjugated hyperbilirubinemia over both time periods, though the effect was slightly attenuated in more recent years (Table S12). Discussion Obesity at diagnosis is consistently reported as a risk factor for TAH, but the existing literature largely relies on population‐based thresholds for defining obesity, which might not accurately capture TAH risk in children with ALL [6, 9, 11, 13]. We performed a retrospective chart review to examine the relationship between BMI at diagnosis and weight change during the initial phase of ALL chemotherapy and the likelihood of subsequent hepatotoxicity. Our findings are largely consistent with the published literature, suggesting the likelihood of hepatotoxicity increases with increasing BMI but providing new insight on the dose–response relationship between body mass and liver toxicity. Additionally, we present new findings highlighting the potential importance of weight trajectories early in treatment on hepatotoxicity during post‐induction phases of therapy. Obesity and/or overweight at diagnosis, defined as an age‐ and sex‐specific BMI z‐score > 1.65 (obese) or BMI z‐score > 1.0 (overweight), has often been cited as a risk factor for TAH in children with ALL [9, 11, 13]. This is the first study to assess outcome‐specific BMI cut points, indicating that BMI z‐score inflection points for TAH likelihood likely occur between traditional overweight (z = 1.0) and obese (z = 1.65) cut points, with a second cut point just above the traditional cut off for underweight (z = −1.65) for transaminitis. Specifically, we observed evidence that being underweight (BMI z‐score ≤ −1.5) may decrease the hazard of transaminitis, particularly in older children, while being overweight (BMI ≥ 1.5) may increase the hazard of both transaminitis and conjugated hyperbilirubinemia. As age is independently associated with both weight loss during induction and hepatotoxicity risk, residual confounding by age cannot be fully excluded, even after adjustment. However, the consistency of BMI associations across age‐stratified models suggests these findings are not solely attributable to age. Weight changes throughout treatment in children with ALL are well documented, with the most substantial changes occurring during the initial induction phase [16, 28, 29]. When stratifying weight trajectories by diagnostic BMI z‐score threshold, our findings were consistent with recent reports that found children with a higher BMI at diagnosis are more likely to lose weight during induction, while healthy weight and underweight children are more likely to gain weight during this time [30]. In this study, children with a BMI z‐score < 1.5 gained an average of 7% of their baseline weight during induction therapy and largely maintained this throughout treatment, never falling below their diagnostic weight. In contrast, children with a BMI z‐score ≥ 1.5 at diagnosis lost an average of 2.25% of their diagnosis weight during the first phase of treatment and remained below their diagnosis weight through the beginning of maintenance therapy. To our knowledge, the association between treatment‐associated weight change and TAH has not been described. While transaminitis was not associated with weight change during induction therapy independent of diagnostic weight group, weight change was associated with subsequent elevations in post‐induction conjugated bilirubin. While mild elevations in conjugated bilirubin are expected, particularly during asparaginase therapy, the threshold used to define conjugated hyperbilirubinemia in this study was selected to reflect severe, dose‐modifying toxicity. Previous research has indicated that dose‐modifying conjugated hyperbilirubinemia is associated with poorer overall and event‐free survival, highlighting the importance of examining this outcome. In this study, for each percentage of weight loss during induction, the hazard of post‐induction conjugated hyperbilirubinemia increased by 8%. While the per unit hazard ratio for weight loss is modest, the clinical relevance becomes apparent when considered cumulatively. For the 20% of children who lost ≥ 5% of their diagnostic weight during induction, there was an over 40% increase in hazard of conjugated hyperbilirubinemia. Notably, after accounting for weight change during induction therapy, diagnostic BMI class was no longer a statistically significant predictor of conjugated hyperbilirubinemia in this population. Analyses stratified by age and diagnostic year demonstrated that the increase in hazard associated with weight loss during induction was consistent across age groups and time periods. Several biological mechanisms could plausibly link diagnostic BMI and TAH. Children with elevated BMI are more likely to have excess adiposity and potentially hepatic steatosis at ALL diagnosis, which may increase vulnerability to hepatotoxic agents used throughout ALL treatment. This is particularly relevant for Hispanic children, who have some of the highest rates of hepatic steatosis of any racial or ethnic group [31], consistent with the association observed between Hispanic ethnicity and transaminitis in this study. While abdominal imaging data was not systematically available to assess hepatic steatosis directly, this represents a biologically plausible mechanism warranting prospective investigation. Notably, 75% of post‐induction conjugated hyperbilirubinemia events and 65% of post‐induction transaminitis events occurred during maintenance therapy, when asparaginase is no longer administered. While asparaginase is a well‐known hepatotoxic agent whose effects may be amplified by preexisting steatosis or metabolic syndrome [32], the predominance of maintenance‐phase events suggests that hepatotoxicity in this cohort is not primarily asparaginase‐driven. Hepatotoxicity during maintenance is generally attributed to oral methotrexate and 6‐mercaptopurine, both of which are potentially hepatotoxic [33, 34]. Whether preexisting hepatic steatosis or metabolic dysfunction amplifies the hepatotoxic effects of these agents during maintenance remains an important open question. The underlying biological mechanisms linking weight loss and conjugated hyperbilirubinemia are more difficult to disentangle. Weight loss during induction may be a surrogate marker for ALL disease acuity or severe treatment‐related complications, which may increase vulnerability to later hepatotoxicity. In this case, weight loss is not directly causal but could be a valuable clinical marker to identify children at higher risk of hepatotoxicity for risk stratification or prevention efforts. These findings highlight that the factors contributing to treatment‐associated hyperbilirubinemia and transaminitis might be distinct and raise questions about the potential impact of changes in body composition during induction therapy on future liver toxicity. The reasons some patients experienced weight loss in the current study are not immediately clear and are difficult to ascertain given the retrospective design. While previous studies have reported transient elevations in liver enzymes for some patients following diet‐ or surgery‐induced rapid weight loss [35, 36, 37, 38], these were primarily obese, adult patients who were intentionally losing weight. Pediatric patients being treated for leukemia may be losing weight due to toxicity or illness, rather than positive lifestyle changes. It is also important to note that BMI z‐scores and changes in weight (in kg) do not adequately capture information on patient body composition, particularly in patients with pediatric ALL who tend to experience sarcopenic obesity, or the loss of lean muscle mass with an increase in body fat percentage [39, 40]. Thus, prospective studies are needed to document the causes of weight loss during ALL induction therapy and monitor the impact of changes in body composition on liver enzymes. Strengths of this study include its large, ethnically diverse cohort with extensive clinical data, including longitudinal weight measurements. Weight change was examined as both a continuous and categorical variable, allowing for the evaluation of trends and potential threshold effects. While this study has many strengths, there are also some limitations to note. First, the retrospective nature of this study did not allow for real‐time monitoring of weight changes or the collection of additional detailed biological or clinical data not routinely documented in medical records. As a result, the associations between weight status, weight change and subsequent TAH observed in the current study may be confounded by unmeasured factors or pharmacogenomic features associated with hepatotoxicity. The retrospective design also limited our ability to evaluate potential confounders related to treatment delivery, such as cumulative drug exposure, dose delays or omissions, formulation changes, and/or supportive care practices. Some practices have also changed over time, which cannot be fully accounted for by stratifying on diagnostic year. As a single‐institution retrospective cohort, these findings should be considered hypothesis‐generating pending replication in independent multicenter cohorts. Additionally, childhood ALL treatment typically lasts 2–3 years and many potentially hepatotoxic agents are used, often in combination. Thus, it is difficult to attribute cases of TAH to any single agent. Finally, the number of severe TAH events was relatively small, limiting our statistical power for some analyses. This work adds valuable context to the association between BMI and hazard of hepatotoxicity in children with ALL with potential clinical implications for risk stratification, hepatotoxicity monitoring, and targeted intervention strategies. Prospective research is needed to confirm our finding that weight change during the early phases of childhood ALL treatment impacts later TAH, while the collection of additional biological and clinical data could provide mechanistic insight into the observed associations. Ultimately, this line of inquiry could provide clarity on whether and how excess adiposity and changes in body composition during ALL therapy alter the risk of hepatotoxicity in children with ALL. This could assist in the identification of high‐risk children that could benefit from increased monitoring and prevention efforts [20]. This is particularly relevant in light of increasing evidence suggesting that levocarnitine prophylaxis may reduce the risk of hepatotoxicity [5, 41, 42, 43, 44, 45, 46]. The identification of protective agents in tandem with the recognition of high‐risk features is a key component of targeted prevention efforts and precision medicine strategies to improve outcomes for children with ALL. Author Contributions Emily J. Mason: conceptualization, writing – review and editing, methodology, writing – original draft, data curation, formal analysis, visualization. Joanna S. Yi: writing – review and editing, investigation. Jeremy M. Schraw: writing – review and editing, methodology. John P. Woodhouse: writing – review and editing, data curation. Philip J. Lupo: writing – review and editing, investigation, funding acquisition. Michael E. Scheurer: writing – review and editing, investigation, funding acquisition. Mark C. Zobeck: writing – review and editing, methodology. Olga A. Taylor: writing – review and editing, investigation. Melissa B. Harrell: writing – review and editing, data curation. Van Huynh: conceptualization, writing – review and editing. Austin Brown: conceptualization, writing – review and editing, methodology, data curation, funding acquisition, writing – original draft, supervision. Karen R. Rabin: writing – review and editing, funding acquisition. Baojiang Chen: writing – review and editing, data curation. Etan Orgel: conceptualization, writing – review and editing. Steven D. Mittelman: conceptualization, writing – review and editing. M. Monica Gramatges: writing – review and editing, funding acquisition. Funding This work was supported in part by the Cancer Prevention & Research Institute of Texas (CPRIT; RP210064 and RP160771 to M Scheurer; RP230026 to A Brown), the National Cancer Institute (P20CA262733 to K Rabin and P Lupo; U54CA302464 to MM Gramatges, K Rabin, and P Lupo; and UH3CA260607 to M Scheurer, MM Gramatges), and St. Baldrick's Foundation Consortium Research Grant 522277 with support from the Micaela's Army Foundation (P Lupo and K Rabin). Ethics Statement The study protocol (H‐40889) was reviewed and approved by the Institutional Review Board at Baylor College of Medicine and a waiver of informed consent was granted for this retrospective chart review. This research was performed in accordance with the Declaration of Helsinki. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: Cut point analysis for transaminitis(A) and conjugated hyperbilirubinemia(B). Table S1: Frequency of ALL protocols in pediatric ALL cohort. Table S2: Cox regression model of baseline BMI and transaminitis stratified by age. Table S3: Cox regression model of baseline BMI and conjugated hyperbilirubinemia stratified by age. Table S4: Cox regression model of induction weight change (%) and post‐induction transaminitis. Table S5: Cox regression model of induction weight change (%) and post‐induction transaminitis stratified by age. Table S6: Cox regression model of categorical induction weight change (%) and post‐induction transaminitis. Table S7: Cox regression model of induction weight change (%) and post‐induction conjugated hyperbilirubinemia stratified by age. Table S8: Cox regression model of categorical induction weight change (%) and post‐induction conjugated hyperbilirubinemia. Table S9: Cox regression model of baseline BMI and transaminitis stratified by diagnostic year. Table S10: Cox regression model of induction weight change (%) and post‐induction transaminitis stratified by diagnostic year. Table S11: Cox regression model of baseline BMI and conjugated hyperbilirubinemia stratified by diagnostic year. Table S12: Cox regression model of induction weight change (%) and post‐induction conjugated hyperbilirubinemia stratified by diagnostic year.